>Ok, i get it. That's the solution i meant in my last topic, but i'm sure if this could change my overall result...

Just a suggestion of what I'd do. I'd create a function for the calculation, then use proof by mathematical induction to verify the results. I'd generally stick to prime numbers (unless, of course, floating points are possible), beginning with 7, within the range of possible values within the dataset. If the results are consistently accurate within the range of acceptable tolerances, then you shouldn't have any problems.